Authors: Balakumar Sundaralingam, Tucker Hermans
This paper proposes a novel approach to performingin-grasp manipulation planning: the problem of moving an objectfrom an initial pose to a goal pose without breaking or makingcontacts. Our method to perform in-grasp manipulation useskinematic trajectory optimization which requires no knowledgeof dynamic properties of the object or the robot. We define a costfunction that attempts to maintain the initial grasp points, whilerelaxing the constraint that the contacts between finger and objectremain rigid. Hence, we name this new formulation relaxed-rigidity constraints. We implement our approach on an Allegrorobot hand and perform experiments on 10 objects from the YCBdataset. However, the implementation would work for any objectthe robot can grasp. We perform thorough analysis and compareto alternative optimization formulations. Our method reaches thedesired object pose with a median position error of 13mm acrossall of the 500 trials without ever dropping the object.